There is considerable interest on the part of patients and healthcare providers in the development of low cost, long-acting, “user-friendly” protein therapeutics. Proteins are expensive to manufacture and unlike conventional small molecule drugs, are usually not readily absorbed by the body. Moreover they are digested if taken orally. Therefore, proteins must typically be administered by injection. After injection most proteins are cleared rapidly from the body, necessitating frequent, often daily, injections. Patients dislike injections, which leads to reduced compliance and reduced drug efficacy. Some proteins such as erythropoietin (EPO) are effective when administered less often (three times per week for EPO) but this is due to the fact that the proteins are glycosylated. Glycosylation requires that the recombinant proteins be manufactured using mammalian cell expression systems, which is expensive and increases the cost of protein pharmaceuticals.
Thus, there is a strong need to develop protein delivery technologies that lower the costs of protein therapeutics to patients and healthcare providers. One solution to this problem is the development of methods to prolong the circulating half-lives of protein therapeutics in the body so that the proteins do not have to be injected frequently. This solution also satisfies the needs and desires of patients for protein therapeutics that are “user-friendly”, i.e., protein therapeutics that do not require frequent injections.
Covalent modification of proteins with polyethylene glycol (PEG) has proven to be a useful method to extend the circulating half-lives of proteins in the body (Abuchowski et al., 1984; Hershfield, 1987; Meyers et al., 1991). Covalent attachment of PEG to a protein increases the protein's effective size and reduces its rate of clearance from the body. PEGs are commercially available in several sizes, allowing the circulating half-lives of PEG-modified proteins to be tailored for individual indications through use of different size PEGs. Other documented in vivo benefits of PEG modification are an increase in protein solubility, stability (possibly due to protection of the protein from proteases) and a decrease in protein immunogenicity (Katre et al., 1987; Katre, 1990).
One known method for PEGylating proteins uses compounds such as N-hydroxy succinimide (NHS)-PEG to attach PEG to free amines, typically at lysine residues or at the N-terminal amino acid. A major limitation of this approach is that proteins typically contain several lysines, in addition to the N-terminal amino acid, and the PEG moiety attaches to the protein non-specifically at any of the available free amines, resulting in a heterogeneous product mixture. Many NHS-PEGylated proteins are unsuitable for commercial use because of low specific activities and heterogeneity. Inactivation results from covalent modification of one or more lysine residues or the N-terminal amino acid required for biological activity or from covalent attachment of the PEG moiety near the active site of the protein.
Of particular relevance to this application is the finding that modification of human growth hormone (hGH) with amine-reactive reagents, including NHS-PEG reagents, reduces biological activity of the protein by more than 10-fold (Teh and Chapman, 1988; Clark et al., 1996). GH is a 22 kDa protein secreted by the pituitary gland. GH stimulates metabolism of bone, cartilage and muscle and is the body's primary hormone for stimulating somatic growth during childhood. Recombinant human GH (rhGH) is used to treat short stature resulting from GH inadequacy, Turner's Syndrome and renal failure in children. GH is not glycosylated and is fully active when produced in bacteria. The protein has a short in vivo half-life and must be administered by daily subcutaneous injection for maximum effectiveness (MacGillivray et al., 1996).
There is considerable interest in the development of long-acting forms of hGH. Attempts to create long-acting forms of hGH by PEGylating the protein with amine-reactive PEG reagents have met with limited success due to significant reductions in bioactivity upon PEGylation. Further, the protein becomes PEGylated at multiple sites (Clark et al., 1996). hGH contains nine lysines, in addition to the N-terminal amino acid. Certain of these lysines are located in regions of the protein known to be critical for receptor binding (Cunningham et al., 1989; Cunningham and Wells, 1989). Modification of these lysine residues significantly reduces receptor binding and bioactivity of the protein (de la Llosa et al., 1985; Martal et al., 1985; Teh and Chapman, 1988; Cunningham and Wells, 1989). hGH is readily modified by NHS-PEG reagents, but biological activity of the NHS-PEG protein is severely compromised, amounting to only 1% of wild type GH biological activity for a GH protein modified with five 5 kDa PEG molecules (Clark et al., 1996). The EC50 for this multiply PEGylated GH protein is 440 ng/ml or approximately 20 nM (Clark et al., 1996). In addition to possessing significantly reduced biological activity, NHS-PEG-hGH is very heterogeneous due to different numbers of PEG molecules attached to the protein and at different amino acid residues, which has an impact on its usefulness as a potential therapeutic. Clark et al. (1996) showed that the circulating half-life of NHS-PEG-hGH in animals is significantly prolonged relative to non-modified GH. Despite possessing significantly reduced in vitro biological activity, NHS-PEG-hGH was effective and could be administered less often than non-modified hGH in a rat GH-deficiency model (Clark et al., 1996). However, high doses of NHS-PEG-hGH (60-180 μg per injection per rat) were required for efficacy in the animal models due to the low specific activity of the modified protein. There is a clear need for better methods to create PEGylated hGH proteins that retain greater bioactivity. There also is a need to develop methods for PEGylating hGH in a way that creates a homogeneous PEG-hGH product.
Biological activities of several other commercially important proteins are significantly reduced by amine-reactive PEG reagents. EPO contains several lysine residues that are critical for bioactivity of the protein (Boissel et al., 1993; Matthews et al., 1996) and modification of lysine residues in EPO results in near complete loss of biological activity (Wojchowski and Caslake, 1989). Covalent modification of alpha-interferon-2 with amine-reactive PEGs results in 40-75% loss of bioactivity (Goodson and Katre, 1990; Karasiewicz et al., 1995). Loss of biological activity is greatest with large (e.g., 10 kDa) PEGs (Karasiewicz et al., 1995). Covalent modification of G-CSF with amine-reactive PEGs results in greater than 60% loss of bioactivity (Tanaka et al., 1991). Extensive modification of IL-2 with amine-reactive PEGs results in greater than 90% loss of bioactivity (Goodson and Katre, 1990).
A second known method for PEGylating proteins covalently attaches PEG to cysteine residues using cysteine-reactive PEGs. A number of highly specific, cysteine-reactive PEGs with different reactive groups (e.g., maleimide, vinylsulfone) and different size PEGs (2-40 kDa) are commercially available. At neutral pH, these PEG reagents selectively attach to “free” cysteine residues, i.e., cysteine residues not involved in disulfide bonds. Cysteine residues in most proteins participate in disulfide bonds and are not available for PEGylation using cysteine-reactive PEGs. Through in vitro mutagenesis using recombinant DNA techniques, additional cysteine residues can be introduced anywhere into the protein. The newly added “free” cysteines can serve as sites for the specific attachment of a PEG molecule using cysteine-reactive PEGs. The added cysteine residue can be a substitution for an existing amino acid in a protein, added preceding the amino-terminus of the protein or after the carboxy-terminus of the protein, or inserted between two amino acids in the protein. Alternatively, one of two cysteines involved in a native disulfide bond may be deleted or substituted with another amino acid, leaving a native cysteine (the cysteine residue in the protein that normally would form a disulfide bond with the deleted or substituted cysteine residue) free and available for chemical modification. Preferably the amino acid substituted for the cysteine would be a neutral amino acid such as serine or alanine. Growth hormone has two disulfide bonds that can be reduced and alkylated with iodoacetimide without impairing biological activity (Bewley et al., (1969). Each of the four cysteines would be reasonable targets for deletion or substitution by another amino acid.
Several naturally-occurring proteins are known to contain one or more “free” cysteine residues. Examples of such naturally-occurring proteins include human Interleukin (IL)-2, beta interferon (Mark et al., 1984), G-CSF (Lu et al., 1989) and basic fibroblast growth factor (Thompson, 1992). IL-2, G-CSF and beta interferon contain an odd number of cysteine residues, whereas basic fibroblast growth factor contains an even number of cysteine residues.
However, expression of recombinant proteins containing free cysteine residues has been problematic due to reactivity of the free sulfhydryl at physiological conditions. Several recombinant proteins containing free cysteines have been expressed as intracellular proteins in bacteria such as E. coli. Examples include natural proteins such as IL-2, beta interferon, G-CSF, basic fibroblast growth factor and engineered cysteine muteins of IL-2 (Goodson and Katre, 1990), IL-3 (Shaw et al., 1992), Tumor Necrosis Factor Binding Protein (Tuma et al., 1995), IGF-I (Cox and McDermott, 1994), IGFBP-1 (Van Den Berg et al., 1997) and protease nexin and related proteins (Braxton, 1998). All of these proteins were insoluble when expressed intracellularly in bacteria. The insoluble proteins could be refolded into their native conformations by performing a series of denaturation, reduction and refolding procedures. These steps add time and cost to the manufacturing process for producing the proteins in bacteria. Improved stability and yields of IL-2 (Mark et al., 1985) and beta interferon (DeChiara et al., 1986) have been obtained by substituting another amino acid, e.g., serine, for the free cysteine residue. It would be preferable to express the recombinant proteins in a soluble, biologically active form to eliminate these extra steps.
One known method of expressing soluble recombinant proteins in bacteria is to secrete them into the periplasmic space or into the media. It is known that certain recombinant proteins such as GH are expressed in a soluble active form when they are secreted into the E. coli periplasm, whereas they are insoluble when expressed intracellularly in E. coli. Secretion is achieved by fusing DNA sequences encoding growth hormone or other proteins of interest to DNA sequences encoding bacterial signal sequences such as those derived from the stII (Fujimoto et al., 1988) and ompA proteins (Ghrayeb et al., 1984). Secretion of recombinant proteins in bacteria is desirable because the natural N-terminus of the recombinant protein can be maintained. Intracellular expression of recombinant proteins requires that an N-terminal methionine be present at the amino-terminus of the recombinant protein. Methionine is not normally present at the amino-terminus of the mature forms of many human proteins. For example, the amino-terminal amino acid of the mature form of human growth hormone is phenylalanine. An amino-terminal methionine must be added to the amino-terminus of a recombinant protein, if a methionine is not present at this position, in order for the protein to be expressed efficiently in bacteria. Typically addition of the amino-terminal methionine is accomplished by adding an ATG methionine codon preceding the DNA sequence encoding the recombinant protein. The added N-terminal methionine often is not removed from the recombinant protein, particularly if the recombinant protein is insoluble. Such is the case with hGH, where the N-terminal methionine is not removed when the protein is expressed intracellularly in E. coli. The added N-terminal methionine creates a “non-natural” protein that potentially can stimulate an immune response in a human. In contrast, there is no added methionine on hGH that is secreted into the periplasmic space using stII (Chang et al., 1987) or ompA (Cheah et al., 1994) signal sequences; the recombinant protein begins with the native amino-terminal amino acid phenylalanine. The native hGH protein sequence is maintained because of bacterial enzymes that cleave the stII-hGH protein (or ompA-hGH protein) between the stII (or ompA) signal sequence and the start of the mature hGH protein. While the periplasmic space is believed to be an oxidizing environment that should promote disulfide bond formation, coexpression of protein disulfide isomerase with bovine pancreatic trypsin inhibitor resulted in a six-fold increase in the yield of correctly folded protein from the E. coli periplasm (Ostermeier et al., (1996). This result would suggest that periplasmic protein folding can at times be inefficient and is in need of improvement for large scale protein production.
hGH has four cysteines that form two disulfides. hGH can be secreted into the E. coli periplasm using stII or ompA signal sequences. The secreted protein is soluble and biologically active (Hsiung et al., 1986). The predominant secreted form of hGH is a monomer with an apparent molecular weight by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) of 22 kDa. Recombinant hGH can be isolated from the periplasmic space by using an osmotic shock procedure (Koshland and Botstein, 1980), which preferentially releases periplasmic, but not intracellular, proteins into the osmotic shock buffer. The released hGH protein is then purified by column chromatography ((Hsiung et al., 1986).
When similar procedures were attempted to secrete hGH variants containing a free cysteine residue (five cysteines; 2N+1), it was discovered that the recombinant hGH variants formed multimers and aggregates when isolated using standard osmotic shock and purification procedures developed for hGH. Very little of the monomeric hGH variant proteins could be detected by non-reduced SDS-PAGE in the osmotic shock lysates or during purification of the proteins by column chromatography.
Alpha interferon (IFN-α2) also contains four cysteine residues that form two disulfide bonds. IFN-α2 can be secreted into the E. coli periplasm using the stII signal sequence (Voss et al., 1994). The secreted protein is soluble and biologically active (Voss et al., 1994). The predominant secreted form of IFN-α2 is a monomer with an apparent molecular weight by SDS-PAGE of 19 kDa. Secreted recombinant IFN-α2 can be purified by column chromatography (Voss et al., 1994).
When similar procedures were attempted to secrete IFN-α2 variants containing a free cysteine residue (five cysteines; 2N+1), it was discovered that the recombinant IFN-α2 variants formed multimers and aggregates when isolated using standard purification procedures developed for IFN-α2. The IFN-α2 variants eluted from the columns very differently than IFN-α2 and very little of the monomeric IFN-α2 variant proteins could be purified using column chromatography procedures developed for IFN-α2.
An alternative method to synthesizing a protein containing a free cysteine residue is to introduce a thiol group into a protein post-translationally via a chemical reaction with succinimidyl 6-[3-2-pyridyldithio)propionamido]hexanoate (LC-SPDP, commercially available from Pierce Chemical Company). LC-SPDP reacts with lysine residues to create a free sulfhydryl group. Chemically cross-linked dimeric EPO was prepared using this reagent in conjunction with a maleimide protein modifying reagent (Sytkowsk et al., 1998). A heterologous mixture of chemically cross-linked EPO proteins was recovered after purification due to non-specific modification of the various lysine residues in EPO. Enhanced pharmacokinetics and in vivo potency of the chemically cross-linked EPO proteins were observed.
Another method that has been used to increase the size of a protein and improve its in vivo potency involves dimerization of the protein using chemical crosslinking reagents. GH is thought to transduce a cellular signal by cross-linking two GH receptors. A GH-GH dimer might facilitate enhanced receptor dimerization and subsequent amplification of the intracellular signal.
Chemically cross-linked dimeric hGH proteins have been described by Mockridge et al. (1998). Using a water soluble cross-linking reagent 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide (EDC), GH was randomly derivatized to give predominantly amide-linked dimers but also amide-linked multimers, depending on the concentration of EDC reagent used. While an increase in in vivo potency was observed, the final protein preparation was heterogeneous due to non-specific reaction of the EDC reagent with various amino acids in the protein, including lysine, aspartic acid and glutamic acid residues and the amino- and carboxy-termini. Injection of such a preparation into humans would be undesirable due to the toxic nature of EDC, potential immunogenic response to the unnatural amide bond formed between the proteins. Generating consistent batches of a purified protein also would be difficult at the manufacturing scale.
Therefore, despite considerable effort, a need still exists for a process for generating homogeneous preparations of long acting recombinant proteins by enhancement of protein molecular weight. A need also for methods that allow secretion and recovery of recombinant proteins containing free cysteine residues in high yield. The present invention satisfies these needs and provides related advantages as well.